JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 115, D22120, doi:10.1029/2010JD014175, 2010
Atmospheric circulation is reflected in precipitation isotope
gradients over the conterminous United States
Zhongfang Liu,1 Gabriel J. Bowen,1,2 and Jeffrey M. Welker3
Received 9 March 2010; revised 24 August 2010; accepted 8 September 2010; published 30 November 2010.
[1] The stable isotopic (d 18O and d 2H) composition of meteoric precipitation integrates
information on the history of water fluxes to, from, and between air masses as they traverse
the continents. The development of new methods relating isotopic data to water cycle
processes will increase our ability to understand this system and its response to natural and
anthropogenic forcing. Here we present an analysis of 2 years (1992–1993) of precipitation
isotope data from 73 sites across the conterminous United States. We focus on patterns
in the spatial precipitation isotope gradients (rate and direction of isotopic change in
space), a metric which we suggest reflects two factors: (1) variation in the balance of
rainout and land‐surface recycling for air masses moving across the continent and (2) the
spatial juxtaposition of air masses carrying moisture of differing origin and rainout history
(e.g., Pacific versus Gulf of Mexico moisture). We demonstrate that the position of
zones of particularly high precipitation isotope slopes correspond to the time‐averaged
position of air mass boundaries and to regions where prevailing moisture transport
trajectories interact with orographic barriers. Differences in the location of high‐slope
zones between winter and summer seasons and between the same seasons in 1992 and
1993 can be related to differences in circulation and weather patterns. These results
suggest new opportunities for the interpretation of precipitation, vapor, and paleoclimate
water isotope data in the context of regional climate dynamics through spatial analysis.
Citation: Liu, Z., G. J. Bowen, and J. M. Welker (2010), Atmospheric circulation is reflected in precipitation isotope gradients
over the conterminous United States, J. Geophys. Res., 115, D22120, doi:10.1029/2010JD014175.
1. Introduction
18
2
[2] The stable isotope (d O and d H) composition of
meteoric precipitation reflects spatial and temporal variation
in climate on a variety of scales. Such variations are
preserved in authigenic minerals and organic compounds
and represent excellent archives for interpreting past continental climate change. In the past decades, numerous
paleorecords have been investigated for reconstructing
Pleistocene/Holocene climate change across the continental
United States, including lacustrine sediments [e.g., Yu and
Eicher, 1998; Anderson et al., 2001; Schwalb and Dean,
2002], speleothems [e.g., Dorale et al., 1992, 1998] and plant
cellulose [e.g., Feng et al., 2007]. The majority of this work
has used isotopic data as an indirect proxy for local climate
variables (e.g., temperature or precipitation amount) based on
modern empirical isotope/climate correlations [e.g., Wright
and Leavitt, 2006].
1
Department of Earth and Atmospheric Sciences, Purdue University,
West Lafayette, Indiana, USA.
2
Purdue Climate Change Research Center, Purdue University, West
Lafayette, Indiana, USA.
3
Environment and Natural Resources Institute, University of Alaska
Anchorage, Anchorage, Alaska, USA.
Copyright 2010 by the American Geophysical Union.
0148‐0227/10/2010JD014175
[3] The isotopic composition of meteoric precipitation
is also closely related to atmospheric circulation patterns
through their effect on the sources and transport paths of
atmospheric vapor across the continents [e.g., Lawrence
et al., 1982; Friedman et al., 2002a]. In many cases, recognition of past changes in atmospheric circulation is critical
to developing dynamical understanding of paleoclimatic
change and its effects on continental environments [e.g.,
Smith and Hollander, 1999]. An increasing number of
studies attempt to relate isotope records to dynamical climate modes and synoptic features of the atmospheric circulation [Amundson et al., 1996; Yu et al., 1997; Smith and
Hollander, 1999; Feng et al., 2007], a potentially powerful
approach for reconstructing the large‐scale state of climate.
Such analyses require a new understanding of the modern
isotopic expression of climate modes and synoptic features in order to support robust interpretations of isotopic
archive data [Baldini et al., 2008].
[4] Information on the relationship between precipitation
isotopic composition and modern atmospheric circulation
in the United States has been derived from short‐term
sampling conducted by individual research groups, long‐
term monitoring data collected by the International Atomic
Energy Agency’s Global Network for Isotopes in Precipitation (GNIP), and, more recently, the national‐scale United
States Network for Isotopes in Precipitation (USNIP) [Welker,
2000; Dutton et al., 2005; Vachon et al., 2007]. Isotopic vari-
D22120
1 of 14
D22120
LIU ET AL.: PRECIPITATION ISOTOPE GRADIENTS
ation across the contiguous United States exhibits strong
continentality, with d2H and d18O values decreasing from the
coasts to the continental interior along dominant circulation
trajectories in response to temperature‐, convective‐ and front‐
driven rainout of 2H‐ and 18O‐ enriched precipitation. Working in the western United States, Ingraham and Taylor [1991]
demonstrated that changes in the rates of isotopic change
with distance along circulation trajectories could be attributed to variation in the relative rates of rainout and resupply
of water to the atmosphere through evapotranspiration. As
a result of differences in rainout rates, differences in the
overland path length, and differences in vapor source‐region
conditions, air masses arriving at a given location via different circulation trajectories often carry moisture with very
different isotopic composition. For example, there are significant differences in the d2H andd18O values of moisture
reaching the Great Basin via transport from the North
Pacific, subtropical Pacific, Gulf of California, and Gulf of
Mexico, with lower values corresponding to precipitation
events derived from the more northerly sources [Friedman
et al., 2002a]. A similar pattern has been observed for winter
precipitation in upstate New York derived from North
Atlantic, central Atlantic, and Arctic air masses, and in this
system the deuterium excess value (d = d2H − 8 × d18O, a
tracer of vapor source region conditions) of precipitation
also varies between air masses that have interacted with the
Great Lakes (high d) and those that have not [Lawrence
et al., 1982; Burnett et al., 2004]. Regional variation in time‐
averaged precipitation isotopic composition in several parts
of the contiguous United States has been attributed in part
to geographic variation in the relative amounts of water
sourced via different atmospheric trajectories, for example,
annual average precipitation isotope ratios are lower at sites
in the southwest United States that do not receive appreciable summer monsoon precipitation, and low isotope
ratios of Vermont precipitation have been associated with
time periods when Arctic‐derived precipitation is relatively
abundant [Friedman et al., 2002b; Sjostrom and Welker,
2009].
[5] Despite the recognition of the widespread influence of
circulation on meteoric water isotope ratios, existing methods for extracting information on circulation patterns from
isotope data are largely limited to the analysis of specific
systems based on empirically observed isotope/circulation
relationships [e.g., Baldini et al., 2008; Yoshimura et al.,
2008; Birks and Edwards, 2009]. Here we introduce and test
the hypothesis that the juxtaposition of moisture with different sources, rainout history, and isotopic composition along
boundaries in the synoptic circulation should be reflected
in the spatial gradient (slope – Dd18O/Ddistance – and
aspect) of precipitation isotopic composition. Because
large‐scale, synoptic data on atmospheric water isotope
ratios do not yet exist, we test the hypothesis by examining
the relationship between time‐averaged (seasonal) circulation
patterns and precipitation isotopic gradients calculated from
two years of monitoring data for the contiguous United
States. Our analysis demonstrates robust correspondence
between the isotopic gradient and atmospheric circulation
that would not be clearly apparent based on raw isotopic data
from individual stations or sampling networks. In contrast to
previous techniques, which are applied to a single region (e.g.,
circulation controls over an ice core site) or assume a circu-
D22120
lation control (e.g., North Atlantic Oscillation) a priori and
then look for its isotopic expression over a large area, this
technique allows different possible circulation controls over
a large area to emerge empirically from the isotope data. As
analytical advances and coordinated monitoring efforts lead
to modern and paleowater proxy data sets that are denser in
spatial and temporal resolution, the gradient analysis introduced here will represent a powerful approach to elucidating
large‐scale processes in the water cycle and their change
over time.
2. Data and Methods
[6] We analyzed data generated as part of the USNIP
(United States Network for Isotopes in Precipitation) program [Welker, 2000] representing more than 3400 precipitation samples collected weekly at 73 sites across the
conterminous United States for 1992 and 1993. The data
include d18O values (reported in d notation and per mil units
relative to Vienna Standard Mean Ocean Water, or VSMOW,
standard) and corresponding precipitation amount. Raw
weekly values were used to calculate precipitation‐amount
weighted isotope ratios for the summer (June, July, and
August) and winter (January and February; December data
were not used because the available isotopic record spanned
January 1992 to December 1993) seasons of each study year.
[7] Interpolated maps of precipitation d18O values were
produced at a resolution of 2.5 arcmin with Ordinary Kriging [Moyeed and Papritz, 2002] in ArcGIS 9.2 using a
spherical semivariogram model with nugget (all interpolations were conducted using the Geostatistical Analyst extension). Gradient and aspect (direction of steepest descent)
were calculated for the resulting fields using the Slope and
Aspect tools in the Spatial Analyst extension. Because
we were interested in the reflection of large‐scale (i.e.,
>100 km) features of the atmospheric circulation in the
observational data, we chose to conduct the interpolation
without the use of ancillary variables such as elevation
[Bowen and Wilkinson, 2002] that would introduce high‐
amplitude variability in the interpolated surface over short
length scales.
[8] We compared the spatial precipitation d18O gradients
with climate fields representing sea level pressure (SLP),
mean surface air temperature and moisture flux obtained
from the NCEP/NCAR reanalysis data set (available at
http://www.cdc.noaa.gov/data/gridded/data.ncep.reanalysis.
html), as well as precipitation amount obtained from Global
Precipitation Climatology Project (GPCP, available at http://
jisao.washington.edu/data/gpcp). Surface temperature spatial slopes were calculated from the reanalysis grids using
the same methods applied to the isotopic grids. We also
provide a direct comparison with air mass distributions using
Spatial Synoptic Classification (SSC) [Kalkstein et al., 1996]
data (available at http://sheridan.geog.kent.edu/ssc.html). The
SSC is a semiautomatic classification scheme that identifies
six different weather types (dry polar, dry temperate, dry
tropical, moist polar, moist temperate and moist tropical)
across the North American continent based on four‐time‐
daily observations of temperature, dew point, wind, pressure
and cloud cover at an individual site. A detailed description
of SSC system is given by Sheridan [2002]. For our analysis, we considered primarily the dry polar, moist moderate,
2 of 14
D22120
LIU ET AL.: PRECIPITATION ISOTOPE GRADIENTS
D22120
Figure 1. Locations of the physiographic provinces (italicized) and states discussed in the text.
The background color field represents land surface elevation (meters) from the United States Geological
Survey (USGS) GTOPO30 digital elevation model (http://eros.usgs.gov/#/Find_Data/Products_and_
Data_Available/gtopo30_info). AZ, Arizona; CA, California; CO, Colorado; IL, Illinois; ND, North
Dakota; NM, New Mexico; NV, Nevada; OK, Oklahoma; SD, South Dakota; TX, Texas.
and moist tropical air mass classes, as these represent the
contrasting influence of polar, midlatitude oceanic, and tropical oceanic air masses most likely to provide vapor of contrasting isotopic composition to the continental interior. As
an additional proxy for water vapor source history, we also
calculated and mapped the spatial distribution of deuterium
excess using the same ordinary kriging method applied to
the isotope ratio data.
3. Climatological Setting
[9] The hydroclimate of the contiguous United States is
dominated by the competing influence of three distinct air
masses that originate over the Pacific Ocean, Gulf of Mexico/
subtropical Atlantic and the Arctic [Bryson and Hare, 1974].
The relative influence of each air mass varies spatially and
temporally. During summer months, monsoonal moisture
from the Gulf of Mexico and subtropical Atlantic dominate
most of the contiguous United States east of the Rocky
Mountain Front and contribute lesser amounts of water to
the southwestern states. Pacific sources contribute moisture
in the Pacific Northwest (North Pacific air masses) and
southwest (Gulf of California/subtropical Pacific air masses), whereas Arctic systems exert only a minor influence
on the northernmost interior states [Bryson and Hare, 1974].
During the winter, southward displacement of the jet stream
and establishment of high pressure over the continental
interior lead to increased dominance of moist Pacific air
masses across the western United States. Anticyclonic flow
around continental high‐pressure systems brings dry Arctic
air masses deep into the middle of the continental United
States, producing frontal storm systems that severely limit
the propagation of Gulf of Mexico and Atlantic moisture
into the continental interior [Maglaras et al., 1995]. During
both seasons, secondary vapor sources such as continental
evapotranspiration [e.g., Ingraham and Taylor, 1991] and
evaporation from the Great Lakes [Gat et al., 1994; Niziol
et al., 1995; Burnett et al., 2004] contribute additional moisture within some regions.
[10] Reanalysis data for 1992 and 1993 demonstrate substantial differences in circulation patterns and hydroclimatology
for these two years. Relative to 1992, the summer of 1993
featured intensified low pressure over the western Great
Plains and a steeper pressure gradient between the subtropical Atlantic and the continental interior (Figures 1 and 2),
leading to unusually strong transport of moisture from the
Gulf of Mexico to the northern Great Plains states (Figures 1
and 3) and correspondingly high summer rainfall in this
region (Figure 4).
[11] For the winter months, there was a substantial difference in the geographic location of the primary high
pressure for the two study years, with a strong high centered
over the Rocky Mountain interior during the winter of 1992
and a more diffuse high‐pressure system situated over the
northern Great Plains during 1993 (Figures 1–3). During
1992, the west‐shifted high blocked westerly storms and
reduced moisture transport from the Pacific to the southwest and western interior, producing a relatively dry winter
in this region (Figures 3 and 4). East of the Rocky Mountains, the lack of a strong high over the Great Plains during
1992 reduced the occurrence of strong frontal storm systems
and cyclonic moisture transport from the Gulf and Atlantic
into the continental interior. This led to lower winter precipitation amounts in the Midwest and a relatively wet winter
in the Gulf Coast, particularly in Texas (Figures 1 and 4). In
1993, higher precipitation amounts extend east from the
central Great Plains to the Atlantic Coast, suggesting stronger
precipitation from frontal systems in this region.
4. Results
4.1. Precipitation d18O Patterns
[12] Precipitation d18O values exhibit significant spatial
autocorrelation for all time periods studied (Moran’s I >
3 of 14
D22120
LIU ET AL.: PRECIPITATION ISOTOPE GRADIENTS
Figure 2. Sea level pressure fields (SLP) for winter (JF) and summer (JJA) of 1992 and 1993, calculated
from the NCAR/NCEP reanalysis data. The contour interval is 1 mbar.
Figure 3. Vertically integrated moisture transport for 1000–700 hPa levels for winter (JF) and summer
(JJA) of 1992 and 1993, calculated from the NCAR/NCEP reanalysis data. The arrow denotes vertically
integrated moisture flux (kg m−1 s−1) and the color fields denote moisture flux divergence (kg m−2 s−1).
4 of 14
D22120
D22120
LIU ET AL.: PRECIPITATION ISOTOPE GRADIENTS
D22120
Figure 4. Mean surface precipitation fields for winter (JF) and summer (JJA) of 1992 and 1993, calculated from Global Precipitation Climatology Project [Adler et al., 2003] monthly precipitation data. The
contour interval is 100 mm.
0.17; Z > 8.9; p < 0.01). The semivariogram model range
varied from 2509 to 4128 km, suggesting that the interpolation was capturing spatial patterns at scales of hundreds to
thousands of kilometers over which patterns of atmospheric
circulation might be expressed. A nonzero nugget value was
fit for each data set, indicating that local‐scale variation
existed in the data that could not be represented in the kriged
surface. Root‐mean‐square error (RMSE) values for the interpolated surface, determined through cross validation within
Geostatistical Analyst, are lower for the summer months
(1.5‰ and 1.9‰ for 1992 and 1993, respectively) and higher
for winter (3.0‰ and 2.8‰). These values are roughly proportional to the standard deviation of station d 18O values for
each study period (2.9‰ and 3.9‰ for summer 1992 and
1993, and 5.4‰ for both winters).
[13] Interpolated amount‐weighted average d 18O values
accurately reproduce the large‐scale patterns described by
the station data, although local deviations between station
and gridded values are apparent in some cases (Figure 5). The
precipitation isotope ratio maps show similar large‐scale spatial patterns for summer (JJA) and winter (JF) of 1992 and
1993, with values decreasing from the low‐latitude, low‐
elevation coastal regions toward high‐latitude, mountains and
inland regions. The d18O values of winter precipitation span
a larger range than summer precipitation, averaging −12.4‰,
with a standard deviation of 4.1‰, in 1992, and −13.0‰,
with a standard deviation of 4.7‰ in 1993. Summer season
values are significantly higher and less variable (mean =
−7.2‰ and s = 2.4‰ in 1992; mean = −7.5‰ and s =
3.3‰ in 1993). During both years, minimum values occur
throughout the northern Rocky Mountains during the summer and from the northern Rockies across the northern Great
Plains in the winter. Maximum values occur along the Gulf
Coast during all time periods.
4.2. Patterns of Precipitation d18O Gradients
[14] Spatial aspect values for precipitation isotope gradients are broadly similar during each time period, with
northward directions on average and deflection toward the
Rocky Mountain interior within the western states (Figure 6).
In contrast, slope values show strong spatial and temporal
variation. The d18O slopes for winter precipitation are larger
(mean = 0.70‰/100 km for 1992 and 0.75‰/100 km for
1993) and more variable (s = 0.24‰/100 km for 1992 and
0.30‰/100 km for 1993) than those for summer (mean =
0.35‰/100 km for 1992 and 0.41‰/100 km for 1993; s =
0.18‰/100 km for 1992 and 0.14‰/100 km for 1993)
(Figure 7). The d 18O slopes reported here are close to
those documented by [Welker, 2000] for a rainout transect
in the Pacific Northwest (1.2‰/100 km) and [Rozanski et
al., 1993] for a transects in Europe‐Eurasia (0.18‰/100
km), and South America (0.30‰/100 km).
[15] The dominant feature in the isotope gradients for the
summer season is a zone of high d 18O slope (large change in
d18O value over a short distance) located in the midcontinent
(Figure 6). This feature is approximately north‐south oriented and centered on the eastern front of the Rocky Mountains, but the details of its geometry and geographic location
differ among the two study years. During 1992, the high‐
slope zone extends in a WSW‐ENE direction across Arizona and New Mexico before curving northward along the
∼103−105th meridian to the Dakotas, and deflecting eastward to Great Lakes. In contrast, in 1993 the zone trends
from the SSE to NNW across west Texas and Oklahoma
5 of 14
D22120
LIU ET AL.: PRECIPITATION ISOTOPE GRADIENTS
D22120
Figure 5. Spatial distribution of precipitation d18O for winter (JF) and summer (JJA) of 1992 and 1993.
Colored points denote d 18O value of individual data collection sites, and color fields denote spatial patterns generated from ordinary kriging interpolation. All values are in ‰ units.
and extends northward along the ∼103–105th meridian
all the way to the Canadian border. During 1992 the d18O
slopes are slightly steeper in the south than the north, and
the reverse condition occurs in 1993. Secondary zones of
moderately high d 18O slopes occur in vicinity of the Great
Lakes and the Cascades.
[16] Winter patterns of precipitation d18O slopes are distinctly different from those seen during summer months. A
Figure 6. Slope and aspect of the precipitation d18O gradient for winter (JF) and summer (JJA) of 1992
and 1993. Background color fields and arrows denote gradient and aspect, respectively. All values are in
units of ‰/100 km.
6 of 14
D22120
LIU ET AL.: PRECIPITATION ISOTOPE GRADIENTS
D22120
Figure 7. Frequency distribution for precipitation d 18O slope values for winter (JF) and summer (JJA) of
1992 and 1993.
zone of high d 18O slopes envelopes the Rocky Mountains
and northern Great Plains in the form of “U”‐shaped band,
extending southeast from the Cascades through Nevada to
New Mexico (Figures 1 and 6). From there, strong d18O
slopes curve northeastward across the Great Plains to the
Great Lakes region. Differences in this pattern are evident
between the two study years in the Sierra Nevada, where
slopes during the winter of 1992 are much lower than in
1993, and in the Great Plains, where the high‐slope zone is
somewhat stronger and positioned farther to the north during
1993. A secondary zone of relative high d18O slope, which
is much stronger in 1992 than in 1993, exists across the
Great Lakes and north Atlantic Coast regions.
[17] We know of no analytical expression for the uncertainty of mapped d 18O gradients presented here. Although
the analysis presented in section 4.1 demonstrates that the
interpolated d 18O values from which the gradients are calculated have significant uncertainty associated with them,
the attribution and interpretation of this error is uncertain
(e.g., as a product of data error, such as sampling and analysis errors, versus a reflection of real variation in d 18O values
over small scales). As a result, error estimation would likely
be best achieved through a nonparametric numerical (e.g.,
Monte Carlo) method, which is beyond the scope of this
manuscript. Although such work should be pursued, we
believe there is strong evidence that the large‐scale patterns
discussed here are robust. All features of the d 18O gradient
we discuss are spatially continuous and defined by data from
many adjacent measurement stations, implying that they are
not artifacts of single errant data values. For each map, a
single spatial semivariogram model was used across the
entire map domain, and visual inspection suggests that the
spatial distribution of gradient values is not directly related
to the spatial distribution (i.e., spacing) of stations, implying
that the observed patterns are not an artifact of interpolation
model. Last, the broad similarity of the observed seasonal
patterns across 2 years of data, analyzed independently,
lends support to the interpretation that the patterns are robust
features of the precipitation isotope climatology.
4.3. Patterns of Precipitation Deuterium Excess
[18] Spatial patterns of deuterium excess (d) values across
the contiguous United States vary both among summer and
winter seasons and among years (Figure 8). In contrast
with precipitation d 18O values and gradients, the d values
show significant but relatively weak spatial autocorrelation
(Moran’s Index > 0.05; Z > 3.6; p < 0.01), except for the
winter of 1993, when d values were not spatially autocorrelated (Moran’s Index = −0.04; Z > −1.3). In the summer, the d values tend to have a sharp division between
the eastern and western United States. For summer of 1992,
d values of less than 8‰ dominate the continent to the north
and west of an arc extending from the northern Great Lakes
to the SW, across the central Rocky Mountains, and into
the Pacific Northwest, with higher values found elsewhere.
In 1993 this line dividing high and low d values is oriented
N‐S along the ∼103–105th meridian, with low d values
found across the western half of the country. Winter precipitation had slightly larger d values on average, and the
spatial distribution of d is comparatively complex. During
all years and seasons, the highest d values (>15‰) occur in
the northeastern states.
5. Discussion
5.1. Orographic Effect
[19] Decreases in precipitation isotope ratios as air masses
are lifted over high topography in the western United States
are well documented [e.g., Ingraham et al., 1991; Dutton
et al., 2005]. In several cases, zones of high‐precipitation
7 of 14
D22120
LIU ET AL.: PRECIPITATION ISOTOPE GRADIENTS
D22120
Figure 8. Spatial distribution of precipitation deuterium excess values for the winter (JF) and summer
(JJA) of 1992 and 1993. Values for individual data collection sites are shown as colored points, and the
background color fields show spatial patterns of d values (except for the winter of 1993, for which d
values were not spatially autocorrelated) interpolated by ordinary kriging. All values are in units of ‰.
isotope slopes identified on our gradient map (Figure 6) are
approximately colocated with known regions of strong orographic rainout (e.g., maxima over the Cascades, summer
maxima in the vicinity of the Rocky Mountain Front),
suggesting that the interaction of high topography with the
prevailing circulation determines the geographic location of
these high‐slope zones. The gradient maps, however, demonstrate that these features are not static, and vary over time
in response to variation in atmospheric circulation. In the
west coast region, the high‐slope zone tends to be stronger
and more clearly defined in the winter season, and during
both seasons the feature is notably weaker across California
and Nevada during 1992 than 1993 (see section 5). In the
summer season, high‐slope values are situated along parts of
the Rocky Mountain Front, but the location of maximum
slope values shifts between more southerly and northerly
positions in 1992 and 1993, respectively. In the eastern United
States, relatively high slopes are observed in the vicinity of
the Appalachian Mountains during some time periods, in
particular during the winter of 1992, which may in part reflect
orographic effects.
5.2. Temperature Effect
[20] Patterns of precipitation isotope ratio variation across
the contiguous United States closely parallel those of temperature, with regions of low temperature being characterized by low d18O values (Figure 5) [Harvey and Welker,
2000; Dutton et al., 2005; Bowen, 2008]. The importance
of temperature‐driven rainout in governing spatial patterns
of isotopic change can be seen through the pair‐wise comparison of seasonal temperatures and precipitation isotope
ratios for monitoring stations (Figure 9). Isotopic differences
between stations separated by 1500 km or less are well cor-
related with differences in temperature, with slope values
of 0.38 to 0.64‰ °C −1 and R 2 values of 0.42 to 0.68. These
d18O/temperature slopes compare well with values of 0.26 to
0.69‰ °C−1 reported for North America, but the R 2 values
are low relative to previous studies (ranging from 0.68 to
0.97) [Joussaume and Jouzel, 1993; Harvey and Welker,
2000].
[21] We also calculated the approximate slopes expected
for these relationships if they were a pure function of temperature‐driven rainout using the Rayleigh equation
R ¼ Ri f ð1Þ ;
where R and Ri are 18O/16O ratio of the remaining water
vapor and its initial value, respectively. The fraction of the
water vapor remaining ( f ) is determined by the change in
saturation vapor pressure of water, with an exponential
dependence on temperature, and a is the temperature‐
dependent equilibrium liquid‐vapor isotope ratio fractionation. In order to approximate the mean conditions for each
of the four study periods, Dd18O/DT slopes were estimated
using a for the average station temperature (Ta) and calculating f as the ratio of saturation vapor pressure at Ta − 15 to
that at Ta. The observed Dd18O/DT slopes are similar to, but
in all cases lower than, the theoretical ones, and together
with large scatter of observed values are consistent with the
large body of literature that has demonstrated the importance
of nontemperature effects, such as variation in moisture
sources, transport trajectories and recycling from the landscape [Rozanski et al., 1992; Fricke and O’Neil, 1999; Kohn
and Welker, 2005] on controlling spatial precipitation isotope ratio distributions.
8 of 14
D22120
LIU ET AL.: PRECIPITATION ISOTOPE GRADIENTS
D22120
Figure 9. Pairwise d18O and temperature difference between precipitation monitoring sites (for all site
pairs separated by <1500 km) for winter (JF) and summer (JJA) of 1992 and 1993. Black lines indicate
the slope of observed Dd 18O/DT relations, and red lines indicate the slope of modeled Dd 18O/DT relations for temperature‐driven Rayleigh fractionation.
[22] Although spatial variation in temperature clearly
influences the large‐scale pattern of precipitation isotope
ratio variation, comparison of the spatial slopes of seasonal
temperature distributions (Figure 10) with the precipitation
isotope ratio gradient maps (Figure 6) suggests that the
pattern of isotope slopes is only weakly related to temperature distributions. Rayleigh theory predicts that the rate
of temperature‐driven isotopic change over space would be
approximately proportional to the rate of temperature change.
In other words, zones of high‐precipitation isotope slopes
should correspond to zones of strong spatial gradients in
temperature. This is clearly so in some cases: for example
high‐precipitation isotope slopes are colocated with strong
temperature gradients in the southern Rockies during most
of the study periods. The relationship is inconsistent, however: high‐precipitation isotope slopes occur across parts of
the Great Plains during all periods despite the absence of
particularly strong temperature gradients. Conversely, strong
temperature gradients in the Pacific Coast States during the
winter of 1992 are not accompanied by high‐precipitation
isotope slopes. Although variation in spatial temperature gradients may influence the pattern of precipitation isotope
slopes, they appear not to be the only, or perhaps even the
primary, control on this pattern.
5.3. Atmospheric Circulation Controls
[23] The d 18O slope fields display a number of relationships with atmospheric pressure fields, precipitation fields
and moisture transport patterns (Figures 2–4) that suggest
zones of high spatial slope may be related to features of the
atmospheric circulation and moisture transport. In particular,
strong zones of high d 18O slope dissect the midcontinent
during all years and seasons in the vicinity of features of the
atmospheric circulation that govern the extent of penetration
of Gulf and Pacific moisture into the continental interior.
We suggest that these high‐slope zones correspond to the
time‐average position of boundaries between regions dominated by moisture from these sources and the minor arctic
and continental sources.
[24] During summer, low‐pressure systems in the midcontinent drive northward transport of moisture from the
Gulf of Mexico and convergence of Gulf‐ and Pacific‐
sourced moisture in the vicinity of the Rocky Mountain
Front (Figure 3). This boundary is approximately colocated
with the dominant summer season zone of high d 18O slopes,
suggesting that the isotope slopes may reflect the spatial
juxtaposition of the relatively 18O‐enriched Gulf moisture
and 18O‐depleted Pacific moisture in this region. Perhaps
the strongest evidence for the relationship between the high
isotope slope zone and the penetration of Gulf moisture
into the midcontinent comes from comparison of summer
season precipitation and air mass frequency patterns for the
summer of 1992 and 1993. In 1993, stronger and more
northward‐situated low pressure increased the penetration
of Gulf‐derived moisture into the upper Midwest and northern Great Plains relative to the summer of 1992, as reflected
9 of 14
D22120
LIU ET AL.: PRECIPITATION ISOTOPE GRADIENTS
D22120
Figure 10. Spatial gradient (slope) fields for mean surface air temperatures in winter (JF) and summer
(JJA) of 1992 and 1993, calculated from the monthly mean NCAR/NCEP reanalysis data. All values are
in units of °C/100 km.
by northward‐displaced contours of moist tropical (MT) air
mass frequency and precipitation‐amount contour in the
Great Plains (Figures 4 and 11). In each year, the high d18O
slope zone approximately parallels the northern and western
extent of MT air mass influence and corresponding high
precipitation amounts, extending northward and westward to
the Canadian border in 1993 and cutting to the northeast
across the northern Great Plains in 1992.
Figure 11. Air mass frequency for winter (DJF) and summer (JJA) of 1992 and 1993, based on the classifications of Kalkstein et al. [1996]. The frequency of the three dominant synoptic air masses are shown:
dry polar (DP, analogous to traditional Continental Polar), moist moderate (MM, associated with overcast,
humid conditions), and moist tropical (MT, analogous to traditional Maritime Tropical). Background
color fields show the precipitation d18O slope values for each time period. The contour interval is 5%,
with contours shown only for the upper range of frequency values for each air mass type.
10 of 14
D22120
LIU ET AL.: PRECIPITATION ISOTOPE GRADIENTS
[25] In contrast to summer season, winter climate over
the contiguous United States features intensified and more
dynamic interaction of moist tropical (MT), moist moderate (MM) and dry polar (DP) air masses (Figure 11). Previous work has demonstrated that precipitation originating
from Arctic‐derived air masses is characterized by very
low d18O values, and the interaction of these systems with
18
O‐enriched vapor from tropical and midlatitude sources
likely contributes to the higher mean and maximum d 18O
slope values during winter relative to summer (Figure 7).
Across the western United States, the influence of DP air
masses was much stronger in 1993 than in 1992, and the
resultant strong gradients in air mass characteristics between
coastal regions and the continental interior are closely matched
by strong wintertime d18O gradients during 1993. During
1992, the relatively westerly position of wintertime high
pressure (Figure 2) reduced the influence of both DP air
masses in the interior and MT and MM air masses along the
coast. As a result, this winter featured a reduction of land‐
falling Pacific storms and relatively dry conditions in coastal
California, reduced precipitation amount gradients between
the coast and interior, and diminished isotope gradients in
this region (Figures 3, 4, and 11).
[26] East of the Rocky Mountains, a narrow and sharp
transition zone between cool, dry continental air (dry polar,
DP, air mass of Kalkstein et al. [1996]) and relatively warm,
humid maritime air (moist moderate and moist tropical, MM
and MT) occurs along the polar front (Figure 11) [Lawrence
et al., 1982]. During the winter season, this boundary coincides with the track of extratropical “Colorado cyclones”
which lead to the convergence of maritime and polar air
producing strong rainout [e.g., Bierly and Winkler, 2001;
Whittaker and Horn, 1984]. The generalized path of these
systems is well documented and corresponds closely to the
geometry of the eastern branch of the high d18O slope trough
observed in our study.
[27] The strength and position of this precipitation isotope
slope feature varies slightly between the two study years,
and is consistent with variation in the atmospheric circulation for these winters. During 1993, winter season high
pressure in the continental interior was situated relatively far
to the east, leading to a strong juxtaposition of DP and MT
air masses along the Colorado cyclone track and promoting
strong frontal precipitation across the central Great Plains
and upper Midwest (Figures 2, 4, and 11). The resultant
zone of high d18O slope values formed a strong and continuous band along the Colorado cyclone track. Due to the
more westerly position of high‐pressure systems during
1992, pressure gradients in the Great Plains and Midwest
were reduced. Strong juxtaposition of DP and MT air
masses was limited to the southern Great Plains and southern
Rocky Mountains, and precipitation amounts in the area
characterized by the Colorado cyclone were relatively low
(Figures 2, 4, and 11). During this winter, the high d18O
slope zone was of reduced magnitude and was discontinuous, with the highest slope values limited to the southern
Rocky Mountain region where DP and MT air masses were
most closely juxtaposed.
[28] During the winter of 1992, an additional zone of high
d18O slopes occurs in the northeastern states, with a secondary branch extending south of the Great Lakes to Illi-
D22120
nois. This feature is noticeably diminished during the winter
of 1993, although a small zone of very high slopes is located
over the northeastern states and a secondary zone of moderately high slopes occurs farther to the south. This difference may also reflect the varying influence of different air
masses and vapor sources throughout the region. The high
d18O slopes in the region in 1992 approximately correspond
to areas where the MM and MT air masses are juxtaposed.
During winter of 1993, this interface is shifted southward
and weaker, corresponding with the zone of moderately high
d18O slopes in middle Atlantic Coast and southern Appalachians. The small zone of very high d18O slopes over the far
northeastern states in 1993 separates areas strongly influenced by the DP and MM air masses.
[29] Another factor that may contribute to high d18O
slopes in this region is the localized influence of the Great
Lakes on precipitation in down‐wind regions (lake effect).
Previous work has demonstrated that evaporation of water
from the Great Lakes is a significant source of water vapor
contributing to winter season precipitation in areas downwind of the lakes and that the resultant precipitation is
characterized by lower d18O values than precipitation from
other vapor sources [Gat et al., 1994; Machavaram and
Krishnamurthy, 1994, 1995; Burnett et al., 2004]. The formation of these lake‐effect events is driven by destabilization of the overlying atmosphere by vertical fluxes of heat
and recycled moisture from the lake surface, and as a result
their occurrence is typically limited to a region within a few
hundred kilometers of the lakes [Niziol et al., 1995; Burnett
et al., 2004]. The position of the northeastern high d 18O
slope zone approximately corresponds with the expected
maximum extent of lake‐effect precipitation, and we suggest
that high d18O gradients in this region may in part reflect the
diminishing influence of the lake system with distance from
the Lakes. Conditions during 1992 were optimal for expression of this contrast: a relatively dry winter with increased
influence of MM air masses (potentially reflecting both
Atlantic‐ and Great Lakes−influenced air masses as suggested by the two loci of high MM air mass frequency shown
in Figure 11), decreased DP air mass influence, and slightly
above average snowfall in lake effect areas (Figures 4 and 11)
[Burnett et al., 2003]. Conversely, the winter of 1993 was relatively wet across the region, with strong influence of DP
air mass, potential for frontal precipitation across the region
and relatively little enhancement of snowfall in lake‐effect
areas. These patterns suggest enhanced regional precipitation of moisture from maritime sources, potentially diluting
the contrast between lake‐effect and non‐lake‐effect areas
and diminishing the d18O slopes across much of the northeastern United States.
5.4. Comparison With Deuterium Excess Pattern
[30] Although the deuterium excess values of precipitation
are controlled by a complex of local and nonlocal processes,
the spatial distribution of d values can serve as a coarse
tracer for water source given the strong influence of source‐
region conditions on the d values of evaporated water [Gat,
1996]. Patterns of d variation across the contiguous United
States in two summers and winter of 1992 are in general
agreement with the arguments presented so far for vapor‐
source controls on the patterns of d18O slopes (Figure 8).
11 of 14
D22120
LIU ET AL.: PRECIPITATION ISOTOPE GRADIENTS
During summer, low d values (<8‰) consistent with Pacific‐
sourced moisture [Kendall and Coplen, 2001] occur to the
north and west of the steep d18O slope band in the Great
Plains, whereas sites to the south and east of the high‐slope
band are characterized by higher d values. Although the
data are noisy and it is difficult to make a strong argument
for robust differences between the summers, the interpolated patterns of d values for 1992 and 1993 suggest that
variation in the spatial patterns may parallel variation in the
orientation of the high d 18O slope bands for these summers.
Winter precipitation d values in 1992 are consistent with a
strong Pacific source influence across the northern interior
of the continent, as suggested based on d18O gradient analysis. During all seasons relatively high values of d in the
northeastern states are consistent with some contribution of
recycled water (e.g., from over‐lake evaporation) to precipitation [Gat et al., 1994].
6. Conclusion
[31] Spatial analysis of a dense network of precipitation
isotope ratio monitoring sites across the contiguous United
States indicates substantial variation in the rate and direction
of change in isotope values over space, the precipitation
isotope gradient, and suggests that the precipitation d18O
slope varies by an order of magnitude. This variation is
spatially patterned and changes over time, with significant
differences demonstrated here between summer and winter
seasons and between equivalent seasons for two consecutive
years. The patterns of d18O gradients likely reflect a complex of factors that govern the moisture sources and rainout
history of atmospheric vapor contributing to precipitation at
different locations, but through analysis we are able to
propose a suite of features of the atmospheric circulation
that appear to influence the most significant features of the
precipitation isotope gradient distributions. In summer, a
north‐south band of steep gradients in the isotope field
along the Rocky Mountain front corresponds approximately
to zone of convergence of Pacific and the Gulf of Mexico air
masses as defined by the strength and position of low‐
pressure systems over the Great Plains. During 1993, when
strong northwestward transport of Gulf moisture brought
flooding to the upper Midwest and northern Great Plains
states, the band of steep d 18O gradients retreated to the north
and west compared to the summer of 1992, when steep
gradients extended across this region. During winter, the
zone of high isotope gradients encloses the Rocky Mountains and northern Great Plains, reflecting the influence of
the polar front and presence of blocking high‐pressure
systems and cyclone tracks across this region. The west‐
shifted position of continental high pressure during the
winter of 1992 corresponded with a reduction in land‐falling
Pacific storms in the southwestern United States and diminished winter season isotope gradients. During 1993, the
wintertime high‐pressure center was situated farther east
over the Great Plains, promoting frontal precipitation and
cyclogenesis in the Great Plains and Midwest, increasing the
cyclonic transport of Gulf and Atlantic moisture to the continental interior, and intensifying the precipitation isotope gradients in the Great Plains and Midwest while diminishing
those surrounding the Great Lakes “lake‐effect” region.
D22120
[32] The recognition that circulation and vapor‐source
patterns are reflected in precipitation isotope gradient fields
at continental scales presents a number of opportunities
for studying the modern atmospheric and land‐atmosphere
components of the water cycle. Water vapor transport and
mixing have been modeled extensively but are extremely
difficult to verify observationally. Moreover, there is significant concern that anthropogenic land use change and
climate change are significantly modifying these processes
[e.g., Betts et al., 2007; Zhang et al., 2007]. Recent advances in technology promise to enable routine regional‐ to
continental‐scale monitoring of atmospheric water vapor
isotopic composition from satellite and ground‐based platforms within the next decade [Helliker and Noone, 2010].
The spatial gradient analysis proposed here represents a
simple method that could be applied to the resultant data to
derive first‐order information about atmospheric and land‐
atmosphere moisture fluxes and test models for these processes, and could easily be adopted as a routine data analysis
method in atmospheric and precipitation isotope monitoring
efforts.
[33] Within the paleoclimate field, isotope archive records
from individual sites have been widely useful in reconstructing time series for local paleoclimate variables, but in
order to better understand past climate and the relationships
between climate dynamics and climate change impacts,
researchers need to be able to reconstruct synoptic and
dynamic features of climate that are defined by their spatial
structure [e.g., Shinker et al., 2006]. Such reconstructions
may be obtained from single‐site records where relationships between local climate and large‐scale features are
especially robust, but a less ambiguous approach would
involve direct reconstruction of the spatial patterns themselves. This study demonstrates relationships between the
spatial structure of precipitation isotopes and aspects of the
modern atmospheric circulation over the contiguous United
States. Our work suggests that changes in these features of
the synoptic circulation, and in likelihood other similar
features occurring elsewhere in the world, could be directly
reconstructed based on spatially distributed networks of
isotope archive records. Until recently the recovery of
archive records required to support interpretations based on
spatial analysis has been prohibitive, but advances in analytical methodology and recent examples from lacustrine
and tree ring research suggest that this may no longer be the
case [e.g., Feng et al., 2007]. If so, spatial analysis of isotope archive data may produce new, more direct records of
past atmospheric circulation and large‐scale climate features, supporting an improved understanding of past climate
states and climate change impacts.
[34] Acknowledgments. This work was provided by U.S. National
Science Foundation grants DBI‐0743543 and EAR‐0602162 to Bowen
and ESH 0080952 to Welker. We would like to thank Wenwen Tung for
her enthusiastic help and the NADP Coordination Office and the NADP
site operators responsible for precipitation collection. This is Purdue Climate Change Research Center paper 1042.
References
Adler, R. F., et al. (2003), The version‐2 global precipitation climatology
project (GPCP) monthly precipitation analysis (1979–present), J. Hydro-
12 of 14
D22120
LIU ET AL.: PRECIPITATION ISOTOPE GRADIENTS
meteorol., 4, 1147–1167, doi:10.1175/1525-7541(2003)004<1147:
TVGPCP>2.0.CO;2.
Amundson, R., et al. (1996), Isotopic evidence for shifts in atmospheric
circulation patterns during the late Quaternary in mid‐North America,
Geology, 24(1), 23–26, doi:10.1130/0091-7613(1996)024<0023:IEFSIA>
2.3.CO;2.
Anderson, L., et al. (2001), Holocene climate inferred from oxygen isotope
ratios in lake sediments, central Brooks Range, Alaska, Quat. Res., 55(3),
313–321, doi:10.1006/qres.2001.2219.
Baldini, L. M., F. McDermott, A. M. Foley, and J. U. L. Baldini (2008),
Spatial variability in the European winter precipitation d 18O‐NAO relationship: Implications for reconstruction NAO‐mode climate variability
in the Holocene, Geophys. Res. Lett., 35, L04709, doi:10.1029/
2007GL032027.
Betts, R. A., et al. (2007), Projected increase in continental runoff due to
plant responses to increasing carbon dioxide, Nature, 448(7157),
1037–1041, doi:10.1038/nature06045.
Bierly, G. D., and J. A. Winkler (2001), A composite analysis of airstreams
within cold‐season Colorado cyclones, Weather Forecast., 16(1), 57–80,
doi:10.1175/1520-0434(2001)016<0057:ACAOAW>2.0.CO;2.
Birks, S. J., and T. W. D. Edwards (2009), Atmospheric circulation controls
on precipitation isotope‐climate relations in western Canada, Tellus,
Ser. B, 61, 566–576, doi:10.1111/j.1600-0889.2009.00423.x.
Bowen, G. J. (2008), Spatial analysis of the intra‐annual variation of precipitation isotope ratios and its climatological corollaries, J. Geophys.
Res., 113, D05113, doi:10.1029/2007JD009295.
Bowen, G. J., and B. Wilkinson (2002), Spatial distribution of d 18O in
meteoric precipitation, Geology, 30(4), 315–318, doi:10.1130/00917613(2002)030<0315:SDOOIM>2.0.CO;2.
Bryson, R. A., and R. K. Hare (1974), The climate of North America, in
Climates of North America, World Surv. Climatol., vol. 11, pp. 1–47,
Elsevier, New York.
Burnett, A. W., et al. (2003), Increasing Great Lake–effect snowfall during
the Twentieth Century: A regional response to global warming?, J. Clim.,
16, 3535‐3542, doi:10.1175/1520-0442(2003)016<3535:IGLSDT>2.0.
CO;2.
Burnett, A. W., H. T. Mullins, and W. P. Patterson (2004), Relationship
between atmospheric circulation and winter precipitation d 18O in central
New York State, Geophys. Res. Lett., 31, L22209, doi:10.1029/
2004GL021089.
Dorale, J. A., et al. (1992), A high‐resolution record of Holocene climate
change in speleothem calcite from Cold Water Cave, northeast Iowa,
Science, 258(5088), 1626–1630, doi:10.1126/science.258.5088.1626.
Dorale, J. A., et al. (1998), Climate and vegetation history of the mid‐
continent from 75 to 25 ka: A speleothem record from Crevice Cave,
Missouri, USA, Science, 282(5395), 1871–1874, doi:10.1126/science.
282.5395.1871.
Dutton, A., et al. (2005), Spatial distribution and seasonal variation in
18 16
O/ O of modern precipitation and river water across the conterminous
USA, Hydrol. Processes, 19(20), 4121–4146, doi:10.1002/hyp.5876.
Feng, X. H., et al. (2007), The changes in North American atmospheric circulation patterns indicated by wood cellulose, Geology, 35(2), 163–166,
doi:10.1130/G22884A.1.
Fricke, H. C., and J. R. O’Neil (1999), The correlation between 18O/16O
ratios of meteoric water and surface temperature: Its use in investigating
terrestrial climate change over geologic time, Earth Planet. Sci. Lett.,
170, 181–196, doi:10.1016/S0012-821X(99)00105-3.
Friedman, I., J. M. Harris, G. I. Smith, and C. A. Johnson (2002a), Stable
isotope composition of waters in the Great Basin, United States: 1. Air‐
mass trajectories, J. Geophys. Res., 107(D19), 4400, doi:10.1029/
2001JD000565.
Friedman, I., G. I. Smith, C. A. Johnson, and R. J. Moscati (2002b), Stable
isotope compositions of waters in the Great Basin, United States: 2. Modern precipitation, J. Geophys. Res., 107(D19), 4401, doi:10.1029/
2001JD000566.
Gat, J. R. (1996), Oxygen and hydrogen isotopes in the hydrologic cycle,
Annu. Rev. Earth Planet. Sci., 24, 225–262.
Gat, J. R., C. J. Bowser, and C. Kendall (1994), The contribution of evaporation from the Great Lakes to the continental atmosphere: Estimate
based on stable isotope data, Geophys. Res. Lett., 21(7), 557–560,
doi:10.1029/94GL00069.
Harvey, F. E., and J. M. Welker (2000), Stable isotopic composition of
precipitation in the semi‐arid north‐central portion of the US Great
Plains, J. Hydrol., 238, 90–109, doi:10.1016/S0022-1694(00)00316-4.
Helliker, B. R., and D. Noone (2010), Novel approaches for monitoring of
water vapor isotope ratios: Plants, lasers and satellites, in Isoscapes,
pp. 71–88, Elsevier, New York.
D22120
Ingraham, N. L., and B. E. Taylor (1991), Light stable isotope systematics of large‐scale hydrologic regimes in California and Nevada, Water
Resour. Res., 27(1), 77–90, doi:10.1029/90WR01708.
Ingraham, N. L., et al. (1991), Stable isotopic study of precipitation and
spring discharge in southern Nevada, J. Hydrol., 125, 243–258,
doi:10.1016/0022-1694(91)90031-C.
Joussaume, S., and J. Jouzel (1993), Paleoclimatic tracers: An investigation
using an atmospheric general circulation model under ice age conditions:
2. Water isotopes, J. Geophys. Res., 98, 2807–2830, doi:10.1029/
92JD01920.
Kalkstein, L. S., et al. (1996), A new spatial synoptic classification: Application to air‐mass analysis, Int. J. Climatol., 16(9), 983–1004,
doi:10.1002/(SICI)1097-0088(199609)16:9<983::AID-JOC61>3.0.
CO;2-N.
Kendall, C., and T. B. Coplen (2001), Distribution of oxygen‐18 and deuterium in river waters across the United States, Hydrol. Processes, 15(7),
1363–1393, doi:10.1002/hyp.217.
Kohn, M. J., and J. M. Welker (2005), On the temperature correlation of
d 18O in modern precipitation, Earth Planet. Sci. Lett., 231(1–2), 87–96,
doi:10.1016/j.epsl.2004.12.004.
Lawrence, J. R., et al. (1982), Storm trajectories in eastern US D/H isotopic
composition of precipitation, Nature, 296(5858), 638–640, doi:10.1038/
296638a0.
Machavaram, M. V., and R. V. Krishnamurthy (1994), Survey of factors
controlling the stable isotope ratios in precipitation in the Great Lakes
region, USA, Isr. J. Earth Sci., 43, 195–202.
Machavaram, M. V., and R. V. Krishnamurthy (1995), Earth surface evaporative process: A case study from the Great Lakes region of the United
States based on deuterium excess in precipitation, Geochim. Cosmochim.
Acta, 59(20), 4279–4283, doi:10.1016/0016-7037(95)00256-Y.
Maglaras, G. J., et al. (1995), Winter weather forecasting throughout the
eastern United States. Part I: An overview, Weather Forecast., 10(1),
5–20, doi:10.1175/1520-0434(1995)010<0005:WWFTTE>2.0.CO;2.
Moyeed, R. A., and A. Papritz (2002), An empirical comparison of kriging methods for nonlinear spatial point prediction, Math. Geol., 34(4),
365–386, doi:10.1023/A:1015085810154.
Niziol, T. A., et al. (1995), Winter weather forecasting throughout the eastern United States. Part IV: Lake effect snow, Weather Forecast., 10(1),
61–77, doi:10.1175/1520-0434(1995)010<0061:WWFTTE>2.0.CO;2.
Rozanski, K., et al. (1992), Relation between long‐term trends of O‐18
isotope composition of precipitation and climate, Science, 258(5084),
981–985, doi:10.1126/science.258.5084.981.
Rozanski, K., L. Araguas‐Araguas, and R. Gonfiantini (1993), Isotopic patterns in modern global precipitation, in Climate Change in Continental
Isotopic Records, Geophys. Monogr. Ser., vol. 78, edited by P. K. Swart
et al., pp. 1–36, AGU, Washington, D. C.
Schwalb, A., and W. E. Dean (2002), Reconstruction of hydrological
changes and response to effective moisture variations from north‐central
USA lake sediments, Quat. Sci. Rev., 21(12–13), 1541–1554,
doi:10.1016/S0277-3791(01)00121-4.
Sheridan, S. C. (2002), The redevelopment of a weather‐type classification
scheme for North America, Int. J. Climatol., 22(1), 51–68, doi:10.1002/
joc.709.
Shinker, J. J., et al. (2006), Synoptic and dynamic climate controls of North
American mid‐continental aridity, Quat. Sci. Rev., 25(13–14), 1401–1417,
doi:10.1016/j.quascirev.2005.12.012.
Sjostrom, D. J., and J. M. Welker (2009), The influence of air mass source
on the seasonal isotopic composition of precipitation, eastern USA, J. Geochem. Explor., 102(3), 103–112, doi:10.1016/j.gexplo.2009.03.001.
Smith, M. A., and D. J. Hollander (1999), Historical linkage between atmospheric circulation patterns and the oxygen isotopic record of sedimentary carbonates from Lake Mendota, Wisconsin, USA, Geology, 27(7),
589–592, doi:10.1130/0091-7613(1999)027<0589:HLBACP>2.3.CO;2.
Vachon, R. W., J. W. C. White, E. Gutmann, and J. M. Welker (2007),
Amount‐weighted annual isotopic (d 18O) values are affected by the seasonality of precipitation: A sensitivity study, Geophys. Res. Lett., 34,
L21707, doi:10.1029/2007GL030547.
Welker, J. M. (2000), Isotopic (d 18O) characteristics of weekly precipitation
collected across the USA: An initial analysis with application to water
source studies, Hydrol. Processes, 14(8), 1449–1464, doi:10.1002/
1099-1085(20000615)14:8<1449::AID-HYP993>3.0.CO;2-7.
Whittaker, L. M., and L. H. Horn (1984), Northern Hemisphere extratropical cyclone activity for four mid‐season months, Int. J. Climatol., 4(3),
297–310, doi:10.1002/joc.3370040307.
Wright, W. E., and S. W. Leavitt (2006), Boundary layer humidity reconstruction for a semiarid location from tree ring cellulose d 18O, J. Geophys. Res., 111, D18105, doi:10.1029/2005JD006806.
13 of 14
D22120
LIU ET AL.: PRECIPITATION ISOTOPE GRADIENTS
Yoshimura, K., M. Kanamitsu, D. Noone, and T. Oki (2008), Historical
isotope simulation using Reanalysis atmospheric data, J. Geophys.
Res., 113, D19108, doi:10.1029/2008JD010074.
Yu, Z., and U. Eicher (1998), Abrupt climate oscillations during the last
deglaciation in central North America, Science, 282(5397), 2235–2238,
doi:10.1126/science.282.5397.2235.
Yu, Z. C., et al. (1997), Middle Holocene dry climate caused by change in
atmospheric circulation patterns: Evidence from lake levels and stable
isotopes, Geology, 25(3), 251–254, doi:10.1130/0091-7613(1997)025<
0251:MHDCCB>2.3.CO;2.
D22120
Zhang, X. B., et al. (2007), Detection of human influence on twentieth‐
century precipitation trends, Nature, 448(7152), 461–464, doi:10.1038/
nature06025.
G. J. Bowen and Z. Liu, Department of Earth and Atmospheric Sciences,
Purdue University, 550 Stadium Mall Dr., West Lafayette, IN 47907, USA.
(gabe@purdue.edu)
J. M. Welker, Environment and Natural Resources Institute, University
of Alaska Anchorage, 707 A St., Anchorage, AL 99501, USA.
14 of 14